4 #include <linux/errno.h>
8 #include <linux/mmdebug.h>
10 #include <linux/bug.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 #include <linux/resource.h>
22 #include <linux/page_ext.h>
26 struct anon_vma_chain;
29 struct writeback_control;
32 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
33 extern unsigned long max_mapnr;
35 static inline void set_max_mapnr(unsigned long limit)
40 static inline void set_max_mapnr(unsigned long limit) { }
43 extern unsigned long totalram_pages;
44 extern void * high_memory;
45 extern int page_cluster;
48 extern int sysctl_legacy_va_layout;
50 #define sysctl_legacy_va_layout 0
54 #include <asm/pgtable.h>
55 #include <asm/processor.h>
58 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
62 * To prevent common memory management code establishing
63 * a zero page mapping on a read fault.
64 * This macro should be defined within <asm/pgtable.h>.
65 * s390 does this to prevent multiplexing of hardware bits
66 * related to the physical page in case of virtualization.
68 #ifndef mm_forbids_zeropage
69 #define mm_forbids_zeropage(X) (0)
72 extern unsigned long sysctl_user_reserve_kbytes;
73 extern unsigned long sysctl_admin_reserve_kbytes;
75 extern int sysctl_overcommit_memory;
76 extern int sysctl_overcommit_ratio;
77 extern unsigned long sysctl_overcommit_kbytes;
79 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
81 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
84 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
86 /* to align the pointer to the (next) page boundary */
87 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
89 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
90 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
93 * Linux kernel virtual memory manager primitives.
94 * The idea being to have a "virtual" mm in the same way
95 * we have a virtual fs - giving a cleaner interface to the
96 * mm details, and allowing different kinds of memory mappings
97 * (from shared memory to executable loading to arbitrary
101 extern struct kmem_cache *vm_area_cachep;
104 extern struct rb_root nommu_region_tree;
105 extern struct rw_semaphore nommu_region_sem;
107 extern unsigned int kobjsize(const void *objp);
111 * vm_flags in vm_area_struct, see mm_types.h.
113 #define VM_NONE 0x00000000
115 #define VM_READ 0x00000001 /* currently active flags */
116 #define VM_WRITE 0x00000002
117 #define VM_EXEC 0x00000004
118 #define VM_SHARED 0x00000008
120 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
121 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
122 #define VM_MAYWRITE 0x00000020
123 #define VM_MAYEXEC 0x00000040
124 #define VM_MAYSHARE 0x00000080
126 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
127 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
128 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
129 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
130 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
132 #define VM_LOCKED 0x00002000
133 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
135 /* Used by sys_madvise() */
136 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
137 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
139 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
140 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
141 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
142 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
143 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
144 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
145 #define VM_ARCH_2 0x02000000
146 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
148 #ifdef CONFIG_MEM_SOFT_DIRTY
149 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
151 # define VM_SOFTDIRTY 0
154 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
155 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
156 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
157 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
159 #if defined(CONFIG_X86)
160 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
161 #elif defined(CONFIG_PPC)
162 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
163 #elif defined(CONFIG_PARISC)
164 # define VM_GROWSUP VM_ARCH_1
165 #elif defined(CONFIG_METAG)
166 # define VM_GROWSUP VM_ARCH_1
167 #elif defined(CONFIG_IA64)
168 # define VM_GROWSUP VM_ARCH_1
169 #elif !defined(CONFIG_MMU)
170 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
173 #if defined(CONFIG_X86)
174 /* MPX specific bounds table or bounds directory */
175 # define VM_MPX VM_ARCH_2
179 # define VM_GROWSUP VM_NONE
182 /* Bits set in the VMA until the stack is in its final location */
183 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
185 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
186 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
189 #ifdef CONFIG_STACK_GROWSUP
190 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
192 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
196 * Special vmas that are non-mergable, non-mlock()able.
197 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
199 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
201 /* This mask defines which mm->def_flags a process can inherit its parent */
202 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
205 * mapping from the currently active vm_flags protection bits (the
206 * low four bits) to a page protection mask..
208 extern pgprot_t protection_map[16];
210 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
211 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
212 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
213 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
214 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
215 #define FAULT_FLAG_TRIED 0x20 /* Second try */
216 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
219 * vm_fault is filled by the the pagefault handler and passed to the vma's
220 * ->fault function. The vma's ->fault is responsible for returning a bitmask
221 * of VM_FAULT_xxx flags that give details about how the fault was handled.
223 * pgoff should be used in favour of virtual_address, if possible.
226 unsigned int flags; /* FAULT_FLAG_xxx flags */
227 pgoff_t pgoff; /* Logical page offset based on vma */
228 void __user *virtual_address; /* Faulting virtual address */
230 struct page *cow_page; /* Handler may choose to COW */
231 struct page *page; /* ->fault handlers should return a
232 * page here, unless VM_FAULT_NOPAGE
233 * is set (which is also implied by
236 /* for ->map_pages() only */
237 pgoff_t max_pgoff; /* map pages for offset from pgoff till
238 * max_pgoff inclusive */
239 pte_t *pte; /* pte entry associated with ->pgoff */
243 * These are the virtual MM functions - opening of an area, closing and
244 * unmapping it (needed to keep files on disk up-to-date etc), pointer
245 * to the functions called when a no-page or a wp-page exception occurs.
247 struct vm_operations_struct {
248 void (*open)(struct vm_area_struct * area);
249 void (*close)(struct vm_area_struct * area);
250 int (*mremap)(struct vm_area_struct * area);
251 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
252 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
254 /* notification that a previously read-only page is about to become
255 * writable, if an error is returned it will cause a SIGBUS */
256 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
258 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
259 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
261 /* called by access_process_vm when get_user_pages() fails, typically
262 * for use by special VMAs that can switch between memory and hardware
264 int (*access)(struct vm_area_struct *vma, unsigned long addr,
265 void *buf, int len, int write);
267 /* Called by the /proc/PID/maps code to ask the vma whether it
268 * has a special name. Returning non-NULL will also cause this
269 * vma to be dumped unconditionally. */
270 const char *(*name)(struct vm_area_struct *vma);
274 * set_policy() op must add a reference to any non-NULL @new mempolicy
275 * to hold the policy upon return. Caller should pass NULL @new to
276 * remove a policy and fall back to surrounding context--i.e. do not
277 * install a MPOL_DEFAULT policy, nor the task or system default
280 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
283 * get_policy() op must add reference [mpol_get()] to any policy at
284 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
285 * in mm/mempolicy.c will do this automatically.
286 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
287 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
288 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
289 * must return NULL--i.e., do not "fallback" to task or system default
292 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
296 * Called by vm_normal_page() for special PTEs to find the
297 * page for @addr. This is useful if the default behavior
298 * (using pte_page()) would not find the correct page.
300 struct page *(*find_special_page)(struct vm_area_struct *vma,
307 #define page_private(page) ((page)->private)
308 #define set_page_private(page, v) ((page)->private = (v))
310 /* It's valid only if the page is free path or free_list */
311 static inline void set_freepage_migratetype(struct page *page, int migratetype)
313 page->index = migratetype;
316 /* It's valid only if the page is free path or free_list */
317 static inline int get_freepage_migratetype(struct page *page)
323 * FIXME: take this include out, include page-flags.h in
324 * files which need it (119 of them)
326 #include <linux/page-flags.h>
327 #include <linux/huge_mm.h>
330 * Methods to modify the page usage count.
332 * What counts for a page usage:
333 * - cache mapping (page->mapping)
334 * - private data (page->private)
335 * - page mapped in a task's page tables, each mapping
336 * is counted separately
338 * Also, many kernel routines increase the page count before a critical
339 * routine so they can be sure the page doesn't go away from under them.
343 * Drop a ref, return true if the refcount fell to zero (the page has no users)
345 static inline int put_page_testzero(struct page *page)
347 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
348 return atomic_dec_and_test(&page->_count);
352 * Try to grab a ref unless the page has a refcount of zero, return false if
354 * This can be called when MMU is off so it must not access
355 * any of the virtual mappings.
357 static inline int get_page_unless_zero(struct page *page)
359 return atomic_inc_not_zero(&page->_count);
363 * Try to drop a ref unless the page has a refcount of one, return false if
365 * This is to make sure that the refcount won't become zero after this drop.
366 * This can be called when MMU is off so it must not access
367 * any of the virtual mappings.
369 static inline int put_page_unless_one(struct page *page)
371 return atomic_add_unless(&page->_count, -1, 1);
374 extern int page_is_ram(unsigned long pfn);
382 int region_intersects(resource_size_t offset, size_t size, const char *type);
384 /* Support for virtually mapped pages */
385 struct page *vmalloc_to_page(const void *addr);
386 unsigned long vmalloc_to_pfn(const void *addr);
389 * Determine if an address is within the vmalloc range
391 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
392 * is no special casing required.
394 static inline int is_vmalloc_addr(const void *x)
397 unsigned long addr = (unsigned long)x;
399 return addr >= VMALLOC_START && addr < VMALLOC_END;
405 extern int is_vmalloc_or_module_addr(const void *x);
407 static inline int is_vmalloc_or_module_addr(const void *x)
413 extern void kvfree(const void *addr);
415 static inline void compound_lock(struct page *page)
417 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
418 VM_BUG_ON_PAGE(PageSlab(page), page);
419 bit_spin_lock(PG_compound_lock, &page->flags);
423 static inline void compound_unlock(struct page *page)
425 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
426 VM_BUG_ON_PAGE(PageSlab(page), page);
427 bit_spin_unlock(PG_compound_lock, &page->flags);
431 static inline unsigned long compound_lock_irqsave(struct page *page)
433 unsigned long uninitialized_var(flags);
434 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
435 local_irq_save(flags);
441 static inline void compound_unlock_irqrestore(struct page *page,
444 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
445 compound_unlock(page);
446 local_irq_restore(flags);
450 static inline struct page *compound_head_by_tail(struct page *tail)
452 struct page *head = tail->first_page;
455 * page->first_page may be a dangling pointer to an old
456 * compound page, so recheck that it is still a tail
457 * page before returning.
460 if (likely(PageTail(tail)))
466 * Since either compound page could be dismantled asynchronously in THP
467 * or we access asynchronously arbitrary positioned struct page, there
468 * would be tail flag race. To handle this race, we should call
469 * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
471 static inline struct page *compound_head(struct page *page)
473 if (unlikely(PageTail(page)))
474 return compound_head_by_tail(page);
479 * If we access compound page synchronously such as access to
480 * allocated page, there is no need to handle tail flag race, so we can
481 * check tail flag directly without any synchronization primitive.
483 static inline struct page *compound_head_fast(struct page *page)
485 if (unlikely(PageTail(page)))
486 return page->first_page;
491 * The atomic page->_mapcount, starts from -1: so that transitions
492 * both from it and to it can be tracked, using atomic_inc_and_test
493 * and atomic_add_negative(-1).
495 static inline void page_mapcount_reset(struct page *page)
497 atomic_set(&(page)->_mapcount, -1);
500 static inline int page_mapcount(struct page *page)
502 VM_BUG_ON_PAGE(PageSlab(page), page);
503 return atomic_read(&page->_mapcount) + 1;
506 static inline int page_count(struct page *page)
508 return atomic_read(&compound_head(page)->_count);
511 static inline bool __compound_tail_refcounted(struct page *page)
513 return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
517 * This takes a head page as parameter and tells if the
518 * tail page reference counting can be skipped.
520 * For this to be safe, PageSlab and PageHeadHuge must remain true on
521 * any given page where they return true here, until all tail pins
522 * have been released.
524 static inline bool compound_tail_refcounted(struct page *page)
526 VM_BUG_ON_PAGE(!PageHead(page), page);
527 return __compound_tail_refcounted(page);
530 static inline void get_huge_page_tail(struct page *page)
533 * __split_huge_page_refcount() cannot run from under us.
535 VM_BUG_ON_PAGE(!PageTail(page), page);
536 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
537 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
538 if (compound_tail_refcounted(page->first_page))
539 atomic_inc(&page->_mapcount);
542 extern bool __get_page_tail(struct page *page);
544 static inline void get_page(struct page *page)
546 if (unlikely(PageTail(page)))
547 if (likely(__get_page_tail(page)))
550 * Getting a normal page or the head of a compound page
551 * requires to already have an elevated page->_count.
553 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
554 atomic_inc(&page->_count);
557 static inline struct page *virt_to_head_page(const void *x)
559 struct page *page = virt_to_page(x);
562 * We don't need to worry about synchronization of tail flag
563 * when we call virt_to_head_page() since it is only called for
564 * already allocated page and this page won't be freed until
565 * this virt_to_head_page() is finished. So use _fast variant.
567 return compound_head_fast(page);
571 * Setup the page count before being freed into the page allocator for
572 * the first time (boot or memory hotplug)
574 static inline void init_page_count(struct page *page)
576 atomic_set(&page->_count, 1);
579 void put_page(struct page *page);
580 void put_pages_list(struct list_head *pages);
582 void split_page(struct page *page, unsigned int order);
583 int split_free_page(struct page *page);
586 * Compound pages have a destructor function. Provide a
587 * prototype for that function and accessor functions.
588 * These are _only_ valid on the head of a PG_compound page.
591 static inline void set_compound_page_dtor(struct page *page,
592 compound_page_dtor *dtor)
594 page[1].compound_dtor = dtor;
597 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
599 return page[1].compound_dtor;
602 static inline int compound_order(struct page *page)
606 return page[1].compound_order;
609 static inline void set_compound_order(struct page *page, unsigned long order)
611 page[1].compound_order = order;
616 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
617 * servicing faults for write access. In the normal case, do always want
618 * pte_mkwrite. But get_user_pages can cause write faults for mappings
619 * that do not have writing enabled, when used by access_process_vm.
621 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
623 if (likely(vma->vm_flags & VM_WRITE))
624 pte = pte_mkwrite(pte);
628 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
629 struct page *page, pte_t *pte, bool write, bool anon);
633 * Multiple processes may "see" the same page. E.g. for untouched
634 * mappings of /dev/null, all processes see the same page full of
635 * zeroes, and text pages of executables and shared libraries have
636 * only one copy in memory, at most, normally.
638 * For the non-reserved pages, page_count(page) denotes a reference count.
639 * page_count() == 0 means the page is free. page->lru is then used for
640 * freelist management in the buddy allocator.
641 * page_count() > 0 means the page has been allocated.
643 * Pages are allocated by the slab allocator in order to provide memory
644 * to kmalloc and kmem_cache_alloc. In this case, the management of the
645 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
646 * unless a particular usage is carefully commented. (the responsibility of
647 * freeing the kmalloc memory is the caller's, of course).
649 * A page may be used by anyone else who does a __get_free_page().
650 * In this case, page_count still tracks the references, and should only
651 * be used through the normal accessor functions. The top bits of page->flags
652 * and page->virtual store page management information, but all other fields
653 * are unused and could be used privately, carefully. The management of this
654 * page is the responsibility of the one who allocated it, and those who have
655 * subsequently been given references to it.
657 * The other pages (we may call them "pagecache pages") are completely
658 * managed by the Linux memory manager: I/O, buffers, swapping etc.
659 * The following discussion applies only to them.
661 * A pagecache page contains an opaque `private' member, which belongs to the
662 * page's address_space. Usually, this is the address of a circular list of
663 * the page's disk buffers. PG_private must be set to tell the VM to call
664 * into the filesystem to release these pages.
666 * A page may belong to an inode's memory mapping. In this case, page->mapping
667 * is the pointer to the inode, and page->index is the file offset of the page,
668 * in units of PAGE_CACHE_SIZE.
670 * If pagecache pages are not associated with an inode, they are said to be
671 * anonymous pages. These may become associated with the swapcache, and in that
672 * case PG_swapcache is set, and page->private is an offset into the swapcache.
674 * In either case (swapcache or inode backed), the pagecache itself holds one
675 * reference to the page. Setting PG_private should also increment the
676 * refcount. The each user mapping also has a reference to the page.
678 * The pagecache pages are stored in a per-mapping radix tree, which is
679 * rooted at mapping->page_tree, and indexed by offset.
680 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
681 * lists, we instead now tag pages as dirty/writeback in the radix tree.
683 * All pagecache pages may be subject to I/O:
684 * - inode pages may need to be read from disk,
685 * - inode pages which have been modified and are MAP_SHARED may need
686 * to be written back to the inode on disk,
687 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
688 * modified may need to be swapped out to swap space and (later) to be read
693 * The zone field is never updated after free_area_init_core()
694 * sets it, so none of the operations on it need to be atomic.
697 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
698 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
699 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
700 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
701 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
704 * Define the bit shifts to access each section. For non-existent
705 * sections we define the shift as 0; that plus a 0 mask ensures
706 * the compiler will optimise away reference to them.
708 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
709 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
710 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
711 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
713 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
714 #ifdef NODE_NOT_IN_PAGE_FLAGS
715 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
716 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
717 SECTIONS_PGOFF : ZONES_PGOFF)
719 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
720 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
721 NODES_PGOFF : ZONES_PGOFF)
724 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
726 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
727 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
730 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
731 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
732 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
733 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
734 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
736 static inline enum zone_type page_zonenum(const struct page *page)
738 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
741 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
742 #define SECTION_IN_PAGE_FLAGS
746 * The identification function is mainly used by the buddy allocator for
747 * determining if two pages could be buddies. We are not really identifying
748 * the zone since we could be using the section number id if we do not have
749 * node id available in page flags.
750 * We only guarantee that it will return the same value for two combinable
753 static inline int page_zone_id(struct page *page)
755 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
758 static inline int zone_to_nid(struct zone *zone)
767 #ifdef NODE_NOT_IN_PAGE_FLAGS
768 extern int page_to_nid(const struct page *page);
770 static inline int page_to_nid(const struct page *page)
772 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
776 #ifdef CONFIG_NUMA_BALANCING
777 static inline int cpu_pid_to_cpupid(int cpu, int pid)
779 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
782 static inline int cpupid_to_pid(int cpupid)
784 return cpupid & LAST__PID_MASK;
787 static inline int cpupid_to_cpu(int cpupid)
789 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
792 static inline int cpupid_to_nid(int cpupid)
794 return cpu_to_node(cpupid_to_cpu(cpupid));
797 static inline bool cpupid_pid_unset(int cpupid)
799 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
802 static inline bool cpupid_cpu_unset(int cpupid)
804 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
807 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
809 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
812 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
813 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
814 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
816 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
819 static inline int page_cpupid_last(struct page *page)
821 return page->_last_cpupid;
823 static inline void page_cpupid_reset_last(struct page *page)
825 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
828 static inline int page_cpupid_last(struct page *page)
830 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
833 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
835 static inline void page_cpupid_reset_last(struct page *page)
837 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
839 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
840 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
842 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
843 #else /* !CONFIG_NUMA_BALANCING */
844 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
846 return page_to_nid(page); /* XXX */
849 static inline int page_cpupid_last(struct page *page)
851 return page_to_nid(page); /* XXX */
854 static inline int cpupid_to_nid(int cpupid)
859 static inline int cpupid_to_pid(int cpupid)
864 static inline int cpupid_to_cpu(int cpupid)
869 static inline int cpu_pid_to_cpupid(int nid, int pid)
874 static inline bool cpupid_pid_unset(int cpupid)
879 static inline void page_cpupid_reset_last(struct page *page)
883 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
887 #endif /* CONFIG_NUMA_BALANCING */
889 static inline struct zone *page_zone(const struct page *page)
891 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
894 #ifdef SECTION_IN_PAGE_FLAGS
895 static inline void set_page_section(struct page *page, unsigned long section)
897 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
898 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
901 static inline unsigned long page_to_section(const struct page *page)
903 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
907 static inline void set_page_zone(struct page *page, enum zone_type zone)
909 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
910 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
913 static inline void set_page_node(struct page *page, unsigned long node)
915 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
916 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
919 static inline void set_page_links(struct page *page, enum zone_type zone,
920 unsigned long node, unsigned long pfn)
922 set_page_zone(page, zone);
923 set_page_node(page, node);
924 #ifdef SECTION_IN_PAGE_FLAGS
925 set_page_section(page, pfn_to_section_nr(pfn));
930 * Some inline functions in vmstat.h depend on page_zone()
932 #include <linux/vmstat.h>
934 static __always_inline void *lowmem_page_address(const struct page *page)
936 return __va(PFN_PHYS(page_to_pfn(page)));
939 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
940 #define HASHED_PAGE_VIRTUAL
943 #if defined(WANT_PAGE_VIRTUAL)
944 static inline void *page_address(const struct page *page)
946 return page->virtual;
948 static inline void set_page_address(struct page *page, void *address)
950 page->virtual = address;
952 #define page_address_init() do { } while(0)
955 #if defined(HASHED_PAGE_VIRTUAL)
956 void *page_address(const struct page *page);
957 void set_page_address(struct page *page, void *virtual);
958 void page_address_init(void);
961 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
962 #define page_address(page) lowmem_page_address(page)
963 #define set_page_address(page, address) do { } while(0)
964 #define page_address_init() do { } while(0)
967 extern void *page_rmapping(struct page *page);
968 extern struct anon_vma *page_anon_vma(struct page *page);
969 extern struct address_space *page_mapping(struct page *page);
971 extern struct address_space *__page_file_mapping(struct page *);
974 struct address_space *page_file_mapping(struct page *page)
976 if (unlikely(PageSwapCache(page)))
977 return __page_file_mapping(page);
979 return page->mapping;
983 * Return the pagecache index of the passed page. Regular pagecache pages
984 * use ->index whereas swapcache pages use ->private
986 static inline pgoff_t page_index(struct page *page)
988 if (unlikely(PageSwapCache(page)))
989 return page_private(page);
993 extern pgoff_t __page_file_index(struct page *page);
996 * Return the file index of the page. Regular pagecache pages use ->index
997 * whereas swapcache pages use swp_offset(->private)
999 static inline pgoff_t page_file_index(struct page *page)
1001 if (unlikely(PageSwapCache(page)))
1002 return __page_file_index(page);
1008 * Return true if this page is mapped into pagetables.
1010 static inline int page_mapped(struct page *page)
1012 return atomic_read(&(page)->_mapcount) >= 0;
1016 * Return true only if the page has been allocated with
1017 * ALLOC_NO_WATERMARKS and the low watermark was not
1018 * met implying that the system is under some pressure.
1020 static inline bool page_is_pfmemalloc(struct page *page)
1023 * Page index cannot be this large so this must be
1024 * a pfmemalloc page.
1026 return page->index == -1UL;
1030 * Only to be called by the page allocator on a freshly allocated
1033 static inline void set_page_pfmemalloc(struct page *page)
1038 static inline void clear_page_pfmemalloc(struct page *page)
1044 * Different kinds of faults, as returned by handle_mm_fault().
1045 * Used to decide whether a process gets delivered SIGBUS or
1046 * just gets major/minor fault counters bumped up.
1049 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1051 #define VM_FAULT_OOM 0x0001
1052 #define VM_FAULT_SIGBUS 0x0002
1053 #define VM_FAULT_MAJOR 0x0004
1054 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1055 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1056 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1057 #define VM_FAULT_SIGSEGV 0x0040
1059 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1060 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1061 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1062 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1064 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1066 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1067 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1070 /* Encode hstate index for a hwpoisoned large page */
1071 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1072 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1075 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1077 extern void pagefault_out_of_memory(void);
1079 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1082 * Flags passed to show_mem() and show_free_areas() to suppress output in
1085 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1087 extern void show_free_areas(unsigned int flags);
1088 extern bool skip_free_areas_node(unsigned int flags, int nid);
1090 int shmem_zero_setup(struct vm_area_struct *);
1092 bool shmem_mapping(struct address_space *mapping);
1094 static inline bool shmem_mapping(struct address_space *mapping)
1100 extern int can_do_mlock(void);
1101 extern int user_shm_lock(size_t, struct user_struct *);
1102 extern void user_shm_unlock(size_t, struct user_struct *);
1105 * Parameter block passed down to zap_pte_range in exceptional cases.
1107 struct zap_details {
1108 struct address_space *check_mapping; /* Check page->mapping if set */
1109 pgoff_t first_index; /* Lowest page->index to unmap */
1110 pgoff_t last_index; /* Highest page->index to unmap */
1113 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1116 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1117 unsigned long size);
1118 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1119 unsigned long size, struct zap_details *);
1120 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1121 unsigned long start, unsigned long end);
1124 * mm_walk - callbacks for walk_page_range
1125 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1126 * this handler is required to be able to handle
1127 * pmd_trans_huge() pmds. They may simply choose to
1128 * split_huge_page() instead of handling it explicitly.
1129 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1130 * @pte_hole: if set, called for each hole at all levels
1131 * @hugetlb_entry: if set, called for each hugetlb entry
1132 * @test_walk: caller specific callback function to determine whether
1133 * we walk over the current vma or not. A positive returned
1134 * value means "do page table walk over the current vma,"
1135 * and a negative one means "abort current page table walk
1136 * right now." 0 means "skip the current vma."
1137 * @mm: mm_struct representing the target process of page table walk
1138 * @vma: vma currently walked (NULL if walking outside vmas)
1139 * @private: private data for callbacks' usage
1141 * (see the comment on walk_page_range() for more details)
1144 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1145 unsigned long next, struct mm_walk *walk);
1146 int (*pte_entry)(pte_t *pte, unsigned long addr,
1147 unsigned long next, struct mm_walk *walk);
1148 int (*pte_hole)(unsigned long addr, unsigned long next,
1149 struct mm_walk *walk);
1150 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1151 unsigned long addr, unsigned long next,
1152 struct mm_walk *walk);
1153 int (*test_walk)(unsigned long addr, unsigned long next,
1154 struct mm_walk *walk);
1155 struct mm_struct *mm;
1156 struct vm_area_struct *vma;
1160 int walk_page_range(unsigned long addr, unsigned long end,
1161 struct mm_walk *walk);
1162 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1163 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1164 unsigned long end, unsigned long floor, unsigned long ceiling);
1165 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1166 struct vm_area_struct *vma);
1167 void unmap_mapping_range(struct address_space *mapping,
1168 loff_t const holebegin, loff_t const holelen, int even_cows);
1169 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1170 unsigned long *pfn);
1171 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1172 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1173 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1174 void *buf, int len, int write);
1176 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1177 loff_t const holebegin, loff_t const holelen)
1179 unmap_mapping_range(mapping, holebegin, holelen, 0);
1182 extern void truncate_pagecache(struct inode *inode, loff_t new);
1183 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1184 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1185 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1186 int truncate_inode_page(struct address_space *mapping, struct page *page);
1187 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1188 int invalidate_inode_page(struct page *page);
1191 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1192 unsigned long address, unsigned int flags);
1193 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1194 unsigned long address, unsigned int fault_flags);
1196 static inline int handle_mm_fault(struct mm_struct *mm,
1197 struct vm_area_struct *vma, unsigned long address,
1200 /* should never happen if there's no MMU */
1202 return VM_FAULT_SIGBUS;
1204 static inline int fixup_user_fault(struct task_struct *tsk,
1205 struct mm_struct *mm, unsigned long address,
1206 unsigned int fault_flags)
1208 /* should never happen if there's no MMU */
1214 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1215 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1216 void *buf, int len, int write);
1218 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1219 unsigned long start, unsigned long nr_pages,
1220 unsigned int foll_flags, struct page **pages,
1221 struct vm_area_struct **vmas, int *nonblocking);
1222 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1223 unsigned long start, unsigned long nr_pages,
1224 int write, int force, struct page **pages,
1225 struct vm_area_struct **vmas);
1226 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1227 unsigned long start, unsigned long nr_pages,
1228 int write, int force, struct page **pages,
1230 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1231 unsigned long start, unsigned long nr_pages,
1232 int write, int force, struct page **pages,
1233 unsigned int gup_flags);
1234 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1235 unsigned long start, unsigned long nr_pages,
1236 int write, int force, struct page **pages);
1237 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1238 struct page **pages);
1240 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1241 struct page **pages);
1242 int get_kernel_page(unsigned long start, int write, struct page **pages);
1243 struct page *get_dump_page(unsigned long addr);
1245 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1246 extern void do_invalidatepage(struct page *page, unsigned int offset,
1247 unsigned int length);
1249 int __set_page_dirty_nobuffers(struct page *page);
1250 int __set_page_dirty_no_writeback(struct page *page);
1251 int redirty_page_for_writepage(struct writeback_control *wbc,
1253 void account_page_dirtied(struct page *page, struct address_space *mapping,
1254 struct mem_cgroup *memcg);
1255 void account_page_cleaned(struct page *page, struct address_space *mapping,
1256 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1257 int set_page_dirty(struct page *page);
1258 int set_page_dirty_lock(struct page *page);
1259 void cancel_dirty_page(struct page *page);
1260 int clear_page_dirty_for_io(struct page *page);
1262 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1264 /* Is the vma a continuation of the stack vma above it? */
1265 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1267 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1270 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1273 return (vma->vm_flags & VM_GROWSDOWN) &&
1274 (vma->vm_start == addr) &&
1275 !vma_growsdown(vma->vm_prev, addr);
1278 /* Is the vma a continuation of the stack vma below it? */
1279 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1281 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1284 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1287 return (vma->vm_flags & VM_GROWSUP) &&
1288 (vma->vm_end == addr) &&
1289 !vma_growsup(vma->vm_next, addr);
1292 extern struct task_struct *task_of_stack(struct task_struct *task,
1293 struct vm_area_struct *vma, bool in_group);
1295 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1296 unsigned long old_addr, struct vm_area_struct *new_vma,
1297 unsigned long new_addr, unsigned long len,
1298 bool need_rmap_locks);
1299 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1300 unsigned long end, pgprot_t newprot,
1301 int dirty_accountable, int prot_numa);
1302 extern int mprotect_fixup(struct vm_area_struct *vma,
1303 struct vm_area_struct **pprev, unsigned long start,
1304 unsigned long end, unsigned long newflags);
1307 * doesn't attempt to fault and will return short.
1309 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1310 struct page **pages);
1312 * per-process(per-mm_struct) statistics.
1314 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1316 long val = atomic_long_read(&mm->rss_stat.count[member]);
1318 #ifdef SPLIT_RSS_COUNTING
1320 * counter is updated in asynchronous manner and may go to minus.
1321 * But it's never be expected number for users.
1326 return (unsigned long)val;
1329 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1331 atomic_long_add(value, &mm->rss_stat.count[member]);
1334 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1336 atomic_long_inc(&mm->rss_stat.count[member]);
1339 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1341 atomic_long_dec(&mm->rss_stat.count[member]);
1344 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1346 return get_mm_counter(mm, MM_FILEPAGES) +
1347 get_mm_counter(mm, MM_ANONPAGES);
1350 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1352 return max(mm->hiwater_rss, get_mm_rss(mm));
1355 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1357 return max(mm->hiwater_vm, mm->total_vm);
1360 static inline void update_hiwater_rss(struct mm_struct *mm)
1362 unsigned long _rss = get_mm_rss(mm);
1364 if ((mm)->hiwater_rss < _rss)
1365 (mm)->hiwater_rss = _rss;
1368 static inline void update_hiwater_vm(struct mm_struct *mm)
1370 if (mm->hiwater_vm < mm->total_vm)
1371 mm->hiwater_vm = mm->total_vm;
1374 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1376 mm->hiwater_rss = get_mm_rss(mm);
1379 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1380 struct mm_struct *mm)
1382 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1384 if (*maxrss < hiwater_rss)
1385 *maxrss = hiwater_rss;
1388 #if defined(SPLIT_RSS_COUNTING)
1389 void sync_mm_rss(struct mm_struct *mm);
1391 static inline void sync_mm_rss(struct mm_struct *mm)
1396 int vma_wants_writenotify(struct vm_area_struct *vma);
1398 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1400 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1404 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1408 #ifdef __PAGETABLE_PUD_FOLDED
1409 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1410 unsigned long address)
1415 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1418 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1419 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1420 unsigned long address)
1425 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1427 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1432 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1433 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1436 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1438 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1440 atomic_long_set(&mm->nr_pmds, 0);
1443 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1445 return atomic_long_read(&mm->nr_pmds);
1448 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1450 atomic_long_inc(&mm->nr_pmds);
1453 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1455 atomic_long_dec(&mm->nr_pmds);
1459 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1460 pmd_t *pmd, unsigned long address);
1461 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1464 * The following ifdef needed to get the 4level-fixup.h header to work.
1465 * Remove it when 4level-fixup.h has been removed.
1467 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1468 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1470 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1471 NULL: pud_offset(pgd, address);
1474 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1476 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1477 NULL: pmd_offset(pud, address);
1479 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1481 #if USE_SPLIT_PTE_PTLOCKS
1482 #if ALLOC_SPLIT_PTLOCKS
1483 void __init ptlock_cache_init(void);
1484 extern bool ptlock_alloc(struct page *page);
1485 extern void ptlock_free(struct page *page);
1487 static inline spinlock_t *ptlock_ptr(struct page *page)
1491 #else /* ALLOC_SPLIT_PTLOCKS */
1492 static inline void ptlock_cache_init(void)
1496 static inline bool ptlock_alloc(struct page *page)
1501 static inline void ptlock_free(struct page *page)
1505 static inline spinlock_t *ptlock_ptr(struct page *page)
1509 #endif /* ALLOC_SPLIT_PTLOCKS */
1511 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1513 return ptlock_ptr(pmd_page(*pmd));
1516 static inline bool ptlock_init(struct page *page)
1519 * prep_new_page() initialize page->private (and therefore page->ptl)
1520 * with 0. Make sure nobody took it in use in between.
1522 * It can happen if arch try to use slab for page table allocation:
1523 * slab code uses page->slab_cache and page->first_page (for tail
1524 * pages), which share storage with page->ptl.
1526 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1527 if (!ptlock_alloc(page))
1529 spin_lock_init(ptlock_ptr(page));
1533 /* Reset page->mapping so free_pages_check won't complain. */
1534 static inline void pte_lock_deinit(struct page *page)
1536 page->mapping = NULL;
1540 #else /* !USE_SPLIT_PTE_PTLOCKS */
1542 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1544 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1546 return &mm->page_table_lock;
1548 static inline void ptlock_cache_init(void) {}
1549 static inline bool ptlock_init(struct page *page) { return true; }
1550 static inline void pte_lock_deinit(struct page *page) {}
1551 #endif /* USE_SPLIT_PTE_PTLOCKS */
1553 static inline void pgtable_init(void)
1555 ptlock_cache_init();
1556 pgtable_cache_init();
1559 static inline bool pgtable_page_ctor(struct page *page)
1561 inc_zone_page_state(page, NR_PAGETABLE);
1562 return ptlock_init(page);
1565 static inline void pgtable_page_dtor(struct page *page)
1567 pte_lock_deinit(page);
1568 dec_zone_page_state(page, NR_PAGETABLE);
1571 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1573 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1574 pte_t *__pte = pte_offset_map(pmd, address); \
1580 #define pte_unmap_unlock(pte, ptl) do { \
1585 #define pte_alloc_map(mm, vma, pmd, address) \
1586 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1588 NULL: pte_offset_map(pmd, address))
1590 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1591 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1593 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1595 #define pte_alloc_kernel(pmd, address) \
1596 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1597 NULL: pte_offset_kernel(pmd, address))
1599 #if USE_SPLIT_PMD_PTLOCKS
1601 static struct page *pmd_to_page(pmd_t *pmd)
1603 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1604 return virt_to_page((void *)((unsigned long) pmd & mask));
1607 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1609 return ptlock_ptr(pmd_to_page(pmd));
1612 static inline bool pgtable_pmd_page_ctor(struct page *page)
1614 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1615 page->pmd_huge_pte = NULL;
1617 return ptlock_init(page);
1620 static inline void pgtable_pmd_page_dtor(struct page *page)
1622 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1623 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1628 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1632 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1634 return &mm->page_table_lock;
1637 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1638 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1640 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1644 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1646 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1651 extern void free_area_init(unsigned long * zones_size);
1652 extern void free_area_init_node(int nid, unsigned long * zones_size,
1653 unsigned long zone_start_pfn, unsigned long *zholes_size);
1654 extern void free_initmem(void);
1657 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1658 * into the buddy system. The freed pages will be poisoned with pattern
1659 * "poison" if it's within range [0, UCHAR_MAX].
1660 * Return pages freed into the buddy system.
1662 extern unsigned long free_reserved_area(void *start, void *end,
1663 int poison, char *s);
1665 #ifdef CONFIG_HIGHMEM
1667 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1668 * and totalram_pages.
1670 extern void free_highmem_page(struct page *page);
1673 extern void adjust_managed_page_count(struct page *page, long count);
1674 extern void mem_init_print_info(const char *str);
1676 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1678 /* Free the reserved page into the buddy system, so it gets managed. */
1679 static inline void __free_reserved_page(struct page *page)
1681 ClearPageReserved(page);
1682 init_page_count(page);
1686 static inline void free_reserved_page(struct page *page)
1688 __free_reserved_page(page);
1689 adjust_managed_page_count(page, 1);
1692 static inline void mark_page_reserved(struct page *page)
1694 SetPageReserved(page);
1695 adjust_managed_page_count(page, -1);
1699 * Default method to free all the __init memory into the buddy system.
1700 * The freed pages will be poisoned with pattern "poison" if it's within
1701 * range [0, UCHAR_MAX].
1702 * Return pages freed into the buddy system.
1704 static inline unsigned long free_initmem_default(int poison)
1706 extern char __init_begin[], __init_end[];
1708 return free_reserved_area(&__init_begin, &__init_end,
1709 poison, "unused kernel");
1712 static inline unsigned long get_num_physpages(void)
1715 unsigned long phys_pages = 0;
1717 for_each_online_node(nid)
1718 phys_pages += node_present_pages(nid);
1723 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1725 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1726 * zones, allocate the backing mem_map and account for memory holes in a more
1727 * architecture independent manner. This is a substitute for creating the
1728 * zone_sizes[] and zholes_size[] arrays and passing them to
1729 * free_area_init_node()
1731 * An architecture is expected to register range of page frames backed by
1732 * physical memory with memblock_add[_node]() before calling
1733 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1734 * usage, an architecture is expected to do something like
1736 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1738 * for_each_valid_physical_page_range()
1739 * memblock_add_node(base, size, nid)
1740 * free_area_init_nodes(max_zone_pfns);
1742 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1743 * registered physical page range. Similarly
1744 * sparse_memory_present_with_active_regions() calls memory_present() for
1745 * each range when SPARSEMEM is enabled.
1747 * See mm/page_alloc.c for more information on each function exposed by
1748 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1750 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1751 unsigned long node_map_pfn_alignment(void);
1752 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1753 unsigned long end_pfn);
1754 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1755 unsigned long end_pfn);
1756 extern void get_pfn_range_for_nid(unsigned int nid,
1757 unsigned long *start_pfn, unsigned long *end_pfn);
1758 extern unsigned long find_min_pfn_with_active_regions(void);
1759 extern void free_bootmem_with_active_regions(int nid,
1760 unsigned long max_low_pfn);
1761 extern void sparse_memory_present_with_active_regions(int nid);
1763 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1765 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1766 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1767 static inline int __early_pfn_to_nid(unsigned long pfn,
1768 struct mminit_pfnnid_cache *state)
1773 /* please see mm/page_alloc.c */
1774 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1775 /* there is a per-arch backend function. */
1776 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1777 struct mminit_pfnnid_cache *state);
1780 extern void set_dma_reserve(unsigned long new_dma_reserve);
1781 extern void memmap_init_zone(unsigned long, int, unsigned long,
1782 unsigned long, enum memmap_context);
1783 extern void setup_per_zone_wmarks(void);
1784 extern int __meminit init_per_zone_wmark_min(void);
1785 extern void mem_init(void);
1786 extern void __init mmap_init(void);
1787 extern void show_mem(unsigned int flags);
1788 extern void si_meminfo(struct sysinfo * val);
1789 extern void si_meminfo_node(struct sysinfo *val, int nid);
1791 extern __printf(3, 4)
1792 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1794 extern void setup_per_cpu_pageset(void);
1796 extern void zone_pcp_update(struct zone *zone);
1797 extern void zone_pcp_reset(struct zone *zone);
1800 extern int min_free_kbytes;
1803 extern atomic_long_t mmap_pages_allocated;
1804 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1806 /* interval_tree.c */
1807 void vma_interval_tree_insert(struct vm_area_struct *node,
1808 struct rb_root *root);
1809 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1810 struct vm_area_struct *prev,
1811 struct rb_root *root);
1812 void vma_interval_tree_remove(struct vm_area_struct *node,
1813 struct rb_root *root);
1814 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1815 unsigned long start, unsigned long last);
1816 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1817 unsigned long start, unsigned long last);
1819 #define vma_interval_tree_foreach(vma, root, start, last) \
1820 for (vma = vma_interval_tree_iter_first(root, start, last); \
1821 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1823 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1824 struct rb_root *root);
1825 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1826 struct rb_root *root);
1827 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1828 struct rb_root *root, unsigned long start, unsigned long last);
1829 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1830 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1831 #ifdef CONFIG_DEBUG_VM_RB
1832 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1835 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1836 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1837 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1840 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1841 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1842 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1843 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1844 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1845 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1846 struct mempolicy *, struct vm_userfaultfd_ctx);
1847 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1848 extern int split_vma(struct mm_struct *,
1849 struct vm_area_struct *, unsigned long addr, int new_below);
1850 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1851 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1852 struct rb_node **, struct rb_node *);
1853 extern void unlink_file_vma(struct vm_area_struct *);
1854 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1855 unsigned long addr, unsigned long len, pgoff_t pgoff,
1856 bool *need_rmap_locks);
1857 extern void exit_mmap(struct mm_struct *);
1859 static inline int check_data_rlimit(unsigned long rlim,
1861 unsigned long start,
1862 unsigned long end_data,
1863 unsigned long start_data)
1865 if (rlim < RLIM_INFINITY) {
1866 if (((new - start) + (end_data - start_data)) > rlim)
1873 extern int mm_take_all_locks(struct mm_struct *mm);
1874 extern void mm_drop_all_locks(struct mm_struct *mm);
1876 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1877 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1879 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1880 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1881 unsigned long addr, unsigned long len,
1882 unsigned long flags,
1883 const struct vm_special_mapping *spec);
1884 /* This is an obsolete alternative to _install_special_mapping. */
1885 extern int install_special_mapping(struct mm_struct *mm,
1886 unsigned long addr, unsigned long len,
1887 unsigned long flags, struct page **pages);
1889 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1891 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1892 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1893 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1894 unsigned long len, unsigned long prot, unsigned long flags,
1895 unsigned long pgoff, unsigned long *populate);
1896 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1899 extern int __mm_populate(unsigned long addr, unsigned long len,
1901 static inline void mm_populate(unsigned long addr, unsigned long len)
1904 (void) __mm_populate(addr, len, 1);
1907 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1910 /* These take the mm semaphore themselves */
1911 extern unsigned long vm_brk(unsigned long, unsigned long);
1912 extern int vm_munmap(unsigned long, size_t);
1913 extern unsigned long vm_mmap(struct file *, unsigned long,
1914 unsigned long, unsigned long,
1915 unsigned long, unsigned long);
1917 struct vm_unmapped_area_info {
1918 #define VM_UNMAPPED_AREA_TOPDOWN 1
1919 unsigned long flags;
1920 unsigned long length;
1921 unsigned long low_limit;
1922 unsigned long high_limit;
1923 unsigned long align_mask;
1924 unsigned long align_offset;
1927 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1928 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1931 * Search for an unmapped address range.
1933 * We are looking for a range that:
1934 * - does not intersect with any VMA;
1935 * - is contained within the [low_limit, high_limit) interval;
1936 * - is at least the desired size.
1937 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1939 static inline unsigned long
1940 vm_unmapped_area(struct vm_unmapped_area_info *info)
1942 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
1943 return unmapped_area_topdown(info);
1945 return unmapped_area(info);
1949 extern void truncate_inode_pages(struct address_space *, loff_t);
1950 extern void truncate_inode_pages_range(struct address_space *,
1951 loff_t lstart, loff_t lend);
1952 extern void truncate_inode_pages_final(struct address_space *);
1954 /* generic vm_area_ops exported for stackable file systems */
1955 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1956 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1957 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1959 /* mm/page-writeback.c */
1960 int write_one_page(struct page *page, int wait);
1961 void task_dirty_inc(struct task_struct *tsk);
1964 #define VM_MAX_READAHEAD 128 /* kbytes */
1965 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1967 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1968 pgoff_t offset, unsigned long nr_to_read);
1970 void page_cache_sync_readahead(struct address_space *mapping,
1971 struct file_ra_state *ra,
1974 unsigned long size);
1976 void page_cache_async_readahead(struct address_space *mapping,
1977 struct file_ra_state *ra,
1981 unsigned long size);
1983 unsigned long max_sane_readahead(unsigned long nr);
1985 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1986 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1988 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1989 extern int expand_downwards(struct vm_area_struct *vma,
1990 unsigned long address);
1992 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1994 #define expand_upwards(vma, address) (0)
1997 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1998 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1999 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2000 struct vm_area_struct **pprev);
2002 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2003 NULL if none. Assume start_addr < end_addr. */
2004 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2006 struct vm_area_struct * vma = find_vma(mm,start_addr);
2008 if (vma && end_addr <= vma->vm_start)
2013 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2015 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2018 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2019 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2020 unsigned long vm_start, unsigned long vm_end)
2022 struct vm_area_struct *vma = find_vma(mm, vm_start);
2024 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2031 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2032 void vma_set_page_prot(struct vm_area_struct *vma);
2034 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2038 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2040 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2044 #ifdef CONFIG_NUMA_BALANCING
2045 unsigned long change_prot_numa(struct vm_area_struct *vma,
2046 unsigned long start, unsigned long end);
2049 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2050 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2051 unsigned long pfn, unsigned long size, pgprot_t);
2052 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2053 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2055 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2057 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2060 struct page *follow_page_mask(struct vm_area_struct *vma,
2061 unsigned long address, unsigned int foll_flags,
2062 unsigned int *page_mask);
2064 static inline struct page *follow_page(struct vm_area_struct *vma,
2065 unsigned long address, unsigned int foll_flags)
2067 unsigned int unused_page_mask;
2068 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2071 #define FOLL_WRITE 0x01 /* check pte is writable */
2072 #define FOLL_TOUCH 0x02 /* mark page accessed */
2073 #define FOLL_GET 0x04 /* do get_page on page */
2074 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2075 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2076 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2077 * and return without waiting upon it */
2078 #define FOLL_POPULATE 0x40 /* fault in page */
2079 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2080 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2081 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2082 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2083 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2085 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2087 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2088 unsigned long size, pte_fn_t fn, void *data);
2090 #ifdef CONFIG_PROC_FS
2091 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2093 static inline void vm_stat_account(struct mm_struct *mm,
2094 unsigned long flags, struct file *file, long pages)
2096 mm->total_vm += pages;
2098 #endif /* CONFIG_PROC_FS */
2100 #ifdef CONFIG_DEBUG_PAGEALLOC
2101 extern bool _debug_pagealloc_enabled;
2102 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2104 static inline bool debug_pagealloc_enabled(void)
2106 return _debug_pagealloc_enabled;
2110 kernel_map_pages(struct page *page, int numpages, int enable)
2112 if (!debug_pagealloc_enabled())
2115 __kernel_map_pages(page, numpages, enable);
2117 #ifdef CONFIG_HIBERNATION
2118 extern bool kernel_page_present(struct page *page);
2119 #endif /* CONFIG_HIBERNATION */
2122 kernel_map_pages(struct page *page, int numpages, int enable) {}
2123 #ifdef CONFIG_HIBERNATION
2124 static inline bool kernel_page_present(struct page *page) { return true; }
2125 #endif /* CONFIG_HIBERNATION */
2128 #ifdef __HAVE_ARCH_GATE_AREA
2129 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2130 extern int in_gate_area_no_mm(unsigned long addr);
2131 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2133 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2137 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2138 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2142 #endif /* __HAVE_ARCH_GATE_AREA */
2144 #ifdef CONFIG_SYSCTL
2145 extern int sysctl_drop_caches;
2146 int drop_caches_sysctl_handler(struct ctl_table *, int,
2147 void __user *, size_t *, loff_t *);
2150 void drop_slab(void);
2151 void drop_slab_node(int nid);
2154 #define randomize_va_space 0
2156 extern int randomize_va_space;
2159 const char * arch_vma_name(struct vm_area_struct *vma);
2160 void print_vma_addr(char *prefix, unsigned long rip);
2162 void sparse_mem_maps_populate_node(struct page **map_map,
2163 unsigned long pnum_begin,
2164 unsigned long pnum_end,
2165 unsigned long map_count,
2168 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2169 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2170 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2171 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2172 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2173 void *vmemmap_alloc_block(unsigned long size, int node);
2174 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2175 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2176 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2178 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2179 void vmemmap_populate_print_last(void);
2180 #ifdef CONFIG_MEMORY_HOTPLUG
2181 void vmemmap_free(unsigned long start, unsigned long end);
2183 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2184 unsigned long size);
2187 MF_COUNT_INCREASED = 1 << 0,
2188 MF_ACTION_REQUIRED = 1 << 1,
2189 MF_MUST_KILL = 1 << 2,
2190 MF_SOFT_OFFLINE = 1 << 3,
2192 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2193 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2194 extern int unpoison_memory(unsigned long pfn);
2195 extern int get_hwpoison_page(struct page *page);
2196 extern int sysctl_memory_failure_early_kill;
2197 extern int sysctl_memory_failure_recovery;
2198 extern void shake_page(struct page *p, int access);
2199 extern atomic_long_t num_poisoned_pages;
2200 extern int soft_offline_page(struct page *page, int flags);
2204 * Error handlers for various types of pages.
2207 MF_IGNORED, /* Error: cannot be handled */
2208 MF_FAILED, /* Error: handling failed */
2209 MF_DELAYED, /* Will be handled later */
2210 MF_RECOVERED, /* Successfully recovered */
2213 enum mf_action_page_type {
2215 MF_MSG_KERNEL_HIGH_ORDER,
2217 MF_MSG_DIFFERENT_COMPOUND,
2218 MF_MSG_POISONED_HUGE,
2221 MF_MSG_UNMAP_FAILED,
2222 MF_MSG_DIRTY_SWAPCACHE,
2223 MF_MSG_CLEAN_SWAPCACHE,
2224 MF_MSG_DIRTY_MLOCKED_LRU,
2225 MF_MSG_CLEAN_MLOCKED_LRU,
2226 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2227 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2230 MF_MSG_TRUNCATED_LRU,
2236 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2237 extern void clear_huge_page(struct page *page,
2239 unsigned int pages_per_huge_page);
2240 extern void copy_user_huge_page(struct page *dst, struct page *src,
2241 unsigned long addr, struct vm_area_struct *vma,
2242 unsigned int pages_per_huge_page);
2243 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2245 extern struct page_ext_operations debug_guardpage_ops;
2246 extern struct page_ext_operations page_poisoning_ops;
2248 #ifdef CONFIG_DEBUG_PAGEALLOC
2249 extern unsigned int _debug_guardpage_minorder;
2250 extern bool _debug_guardpage_enabled;
2252 static inline unsigned int debug_guardpage_minorder(void)
2254 return _debug_guardpage_minorder;
2257 static inline bool debug_guardpage_enabled(void)
2259 return _debug_guardpage_enabled;
2262 static inline bool page_is_guard(struct page *page)
2264 struct page_ext *page_ext;
2266 if (!debug_guardpage_enabled())
2269 page_ext = lookup_page_ext(page);
2270 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2273 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2274 static inline bool debug_guardpage_enabled(void) { return false; }
2275 static inline bool page_is_guard(struct page *page) { return false; }
2276 #endif /* CONFIG_DEBUG_PAGEALLOC */
2278 #if MAX_NUMNODES > 1
2279 void __init setup_nr_node_ids(void);
2281 static inline void setup_nr_node_ids(void) {}
2284 #endif /* __KERNEL__ */
2285 #endif /* _LINUX_MM_H */